Reset log selection at context boundaries instead of clamping on read

[jvanbuel]

Stacked on #695. Both touch clear_state(), so this targets that branch rather than main. GitHub will retarget it to main automatically once #695 merges. Review #695 first.

Two halves of the same problem: log view state outliving the context it belonged to.

1. clear_state() left log state behind

Switching environments cleared logs.all but left logs.current (which try is selected) and logs.scroll_mode pointing into the previous environment's tries. Land on a task with two tries carrying current = 3 from before, and you open on try 2 instead of try 1; a deep Manual { position } scroll leaks into the next log you view.

The codebase already knew how to do this — state/context.rs does exactly the right thing when the task context changes:

if is_new_context {
    self.logs.current = 0;
    self.logs.reset_scroll();
}

clear_state() now uses the same two lines. LogModel::reset_scroll() is documented "used when navigating to a new context" and simply wasn't being called on the environment-switch path.

2. current_index() clamped on read instead of fixing the write

fn current_index(&self) -> usize {
    self.current % self.all.len().max(1)
}

This re-derived a valid index on every single read, papering over a stale current rather than correcting it. update_logs now clamps when it writes the list, so current is always valid and the three readers use it directly. current_index() is gone.

Clamping also beats wrapping on the merits: after the try list shrinks you want the newest try, not whichever one modular arithmetic happens to land on. With current = 4 and three tries, wrapping gives try 2; clamping gives try 3.

Test changes

current_index_wraps_when_selection_exceeds_len encoded the workaround as intended behavior, so it's replaced by two tests on the new contract — one for the clamp, one for the empty list. The existing open_ui_with_empty_logs_does_not_panic regression test still passes; with the modulo gone there is no division left to panic on.

Verification

cargo test --workspace --lib --bins (95 passing), cargo clippy --workspace --all-targets --all-features -- -D warnings, and cargo fmt --all --check all clean.

🤖 Generated with Claude Code

https://claude.ai/code/session_0127jzUgqqy8JVX5RydfQxh8

---

Generated by Claude Code

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on base/target branches other than the default branch.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro Plus

Run ID: 93b7aa14-345a-4735-945d-cf3625229d16

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch claude/reset-log-state-on-boundaries

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}